Robust Estimation of a Genomic Association against the Imbalance among the Multi-class Phenotypes

• 발행기관: 한국자료분석학회
• 저자명: 이재용(Jaeyong Yee) 박미라(Mira Park)
• 간행물 정보: 『Journal of The Korean Data Analysis Society (JKDAS)』Vol.18 No.4, 1741~1750쪽, 전체 10쪽
• 주제분류: 자연과학 > 통계학
• 파일형태: PDF
• 발행일자: 2016.08.30

영문 초록

Balance among the numbers of samples classified by their respective phenotypic classes would affect the results of the genomic association assessments. However it may not be always possible to maintain the balance or there may not be even clear balancing condition, especially with multi-class datasets. It would be desirable to have a method that is able to persistently give a consistent and correct estimation against a certain amount of change or imbalance between the phenotypic classes. We propose to examine the variance of an association measure on an ensemble of confusion matrices formed by assigning the possible combinations of the predicted class to each multifactor genotype without an effort to specify the single best set. Detection power of our method was retained through intentional un-balancing when odd ratio-based estimation showed continuous drop in the detection power. It was found that only 10% of the complete set of predicted class would yield sufficient result. Improvement was obvious and therefore the proposed method may not require the strict balance among the phenotypic classes to assess the correct genomic association.

목차

1. Introduction

2. Methods

3. Analysis

4. Conclusion

References

키워드

genomic association confusion matrix multi-class phenotype imbalance